Summary-based Mendelian Randomization

Date: 2021_03_19

Analyzed by: Sara Saez Atienzar

Background

Mendelian randomization is an analytical method that uses genetic information, in the form of genetic variants (instruments) to infer causal relationships between two traits/phenotypes (known as exposure and outcome, in this case we use MG as outcome). For a good description of the method, see https://elifesciences.org/articles/34408

Important links that will help you to understand the analysis and results can be found in here:

Goal

Perform Mendelian randomization analysis to identify possible causal relationship between traits in ieu-a, ebi-a and ukb-a datasets and MG

Pipeline

Mendelian randomization using the following sofwares/packages (and versions)

MG summary stats GWAS

Jupyter notebook containing the script can be found in here:

Directory and organization

Analysis

ieu-a Rscript

Run as a slurm job all the traits contained in ieu-a

Results

Results from traits in the ukb-a dataset

Select the traits with IVW FDR-corrected p-value < 0.05 and Weighted median and MR Egger row pvalue < 0.05. For the selected traits, get pleitropy and heterogeneity results too. Only two traits are selected for follow-up analysis:

  1. Non-cancer illness code self-reported: hypothyroidism/myxoedema || id:ukb-a-77
  2. Treatment/medication code: levothyroxine sodium || id:ukb-a-190

Results from traits in the ebi-a dataset

Select the traits with IVW FDR-corrected p-value < 0.05 and Weighted median and MR Egger row pvalue < 0.05. For the selected traits, get pleitropy and heterogeneity results too. Three traits are selected for follow-up analysis:

  1. Rheumatoid arthritis || id:ebi-a-GCST002318
  2. Multiple sclerosis || id:ebi-a-GCST005531
  3. Type 1 diabetes || id:ebi-a-GCST005536

Results from traits in the ieu-a dataset

Select the traits with IVW FDR-corrected p-value < 0.05 and Weighted median and MR Egger row pvalue < 0.05. For the selected traits, get pleitropy and heterogeneity results too. NO traits were selected

Follow-up

Plots and radial MR with each of the selected traits

Rheumatoid arthritis || id:ebi-a-GCST002318

Radial MR results

Plots

Multiple sclerosis || id:ebi-a-GCST005531

Radial MR results

Plots

Type 1 diabetes || id:ebi-a-GCST005536

Radial MR

Plots

Non-cancer illness code self-reported: hypothyroidism/myxoedema || id:ukb-a-77

Radial MR

Plots

SIDE NOTE: One of the significant traits is: ukb-a-190, Treatment/medication code: levothyroxine sodium. This medication is used to trait ukb-a-77 so we do NOT need a follow-up for ukb-a-190